AI visibility check
ChatGPT names companies from memory, or from pages it searched a moment ago. You can be in one and missing from the other.
A growth strategist replies by email within two business days. No card required. Or book a call.
Three worth a look, in no order.
You: not named in the answer
Training memory with a cutoff, or the live web when ChatGPT decides a search would help. Consumer users cannot turn the search off.
ChatGPT rewrites the question into its own queries and cites what it read. OpenAI says placement is never assured.
The buyer’s tier, what ChatGPT remembers about them, and their rough location.
Your plan
A week of your buyers’ questions, put to ChatGPT from memory and with the live web. A growth strategist reads every answer.
Inside the plan
1 questionThe category question, named from memory
2 pagesComparison pages cited, you on neither
Finding · one week of answersSituation drop-off
Named from memory for the category question, missing with the live web for every question that names the buyer’s situation.
What the answers lean on
The situation answers cite two comparison pages where the company does not appear.
Could make us wrong
A week of answers is a short read, and comparison pages update on their own schedule.
Next in the plan
The two situation pages the answers lean on, written to the buyer’s question
The read
One tier, one account, one screen is a sample of one. The plan reads the whole shape.
ChatGPT answers from memory or from a search. Memory is what it learned before its cutoff, spring 2026 for the current models. A search reads a handful of pages and cites them.
You can be strong in one and absent from the other. The fix is different for each.
ChatGPT does not search for your buyer’s words. It writes its own queries first, and may write more after the first results.
The companies in that rewrite tend to reach the answer. Yours is on the list or it is not.
Since October 2026, free and Go users get one model and paying users another. A third serves the Pro reasoning option.
Your buyer’s tier decides which ChatGPT names you.
With memory on, ChatGPT may fold what it knows about the person into the search it writes. OpenAI’s own example is a vegan in San Francisco asking for a restaurant.
Two buyers asking the same words can be shown two lists.
A memory answer describes you as you were at the cutoff. An old price or a retired product is said in a calm voice, with no date on the screen.
Being named with a wrong fact is not a win. It is the finding the plan is built to catch.
Since February 2026, free and Go users in some countries see labelled ads below the answer. OpenAI states advertisers cannot shape, rank or alter the answer itself.
The names in the answer are still earned.
The other assistants
Searches when it helps, cites what it read, and keeps memory on by default on personal accounts.
Citations and memory
How Claude decidesThree Google surfaces read one index and name different companies.
Three Google surfaces
How Gemini decidesReads X posts beside the web, by xAI’s own description.
Web and X posts
How Grok decidesThe playbook
A question that contains your name answers about you. We ask what a buyer asks before they have heard of you, in the buyer’s words.
Memory answers and live-web answers are read apart. Missing from one is a footprint problem. Missing from the other is a page problem.
Every mention is checked against what is true today. An old price or a retired product is recorded as what it is, not counted as a win.
Every number carries its date. The same questions are asked again next month, so the change is the finding. Your growth strategist reads the change with you.
You read the plan. An expert did the asking.
Field Notes
Questions
Ask the question your buyer asks before they have heard of you, not your company’s name. Read whether you come back, from memory and with the live web. The Visibility Plan starts there, with a week of your buyers’ questions, and says what to ship next. It starts with an AI visibility check.
Both. It answers from memory unless it decides a search would help. Then it writes its own queries, reads what comes back and cites it. The two answers can name different companies.
Usually one of two reasons. From memory, the web described your competitor more clearly before the cutoff. With the live web, their pages rank for the buyer’s question and yours do not. The plan shows which, so the fix starts in the right place.
It can. With memory on, ChatGPT may fold what it knows about a person into the search it writes for them. Two buyers asking the same words can be shown different lists.
No. ChatGPT shows labelled ads to free and Go users. OpenAI states advertisers cannot shape, rank or alter the answer itself.
Find where the wrong fact lives on the web and correct it at the source. Say the right fact plainly on your own site. Memory answers change when the model does. Live-web answers change when the pages do. The plan tells you which one said it.
ChatGPT, Claude, Gemini, Perplexity, Copilot and Google’s AI Overviews, for a week of your buyers’ questions. Grok is in development. Whether you are named, who is named instead, what is said about you and whether it is right, the pages behind each answer, and what to ship next. Every number dated.
How it starts
